As investors navigate a slower growth, policy-fragmented environment, AI stands out as a long-term theme with the potential to deliver durable growth and broad-based opportunity across the value chain
Volatility has defined equity markets so far this year, with artificial intelligence (AI) stocks at the centre of the action. The pullback in many AI-related names in the first quarter was driven in part by the release of DeepSeek, an advanced model that prompted speculation about a shift in the direction and intensity of the AI infrastructure buildout. At the same time, a mix of other dynamics – including crowding in mega-cap names, systematic selling and sharp factor reversals – added to a broader risk-off tone.
Earnings from major cloud and platform providers have since reaffirmed the enduring strength of the AI story. Firms reported strong cloud demand, persistent capacity constraints and plans to maintain or even increase their capital spending plans. Capital expenditure for AI hyperscalers (Alphabet, Meta, Microsoft and Amazon) is projected to rise to $395 billion in 2027 from $150 billion in 2023, representing a 160% increase (Figure 1). Taken together, these results suggest the AI investment cycle remains very much intact.
Figure 1: To meet surging AI demand, tech giants continue to invest
Capital expenditure ($ billions)
Source: Columbia Threadneedle Investments’ analysis, as of 31 May 2025. Represents historical and forecasted capital expenditures for Alphabet, Meta, Microsoft and Amazon.
More broadly, policy uncertainty and tariffs are creating headwinds that are likely to weigh on growth this year. In this type of environment, investors typically seek companies with more visible, durable earnings streams. Backed by accelerating use cases and falling costs, AI-aligned businesses increasingly fit that profile. As investors acclimatise to a lower visibility macroeconomic environment, AI is poised to reemerge as a focal point for long-term growth.
Adoption is accelerating and deepening
Breakthrough technologies rarely follow a linear adoption path. In fact, the pace of adoption for transformative consumer technologies has increased with each successive wave. Electricity took decades to reach 30% of US households. The internet got there in half the time. With ChatGPT reaching 100 million users within six months, AI appears to be moving even faster than its predecessors.
Importantly, AI adoption is extending well beyond chatbots and search engines into high-impact sectors. This is because AI is not just becoming more capable, it is also becoming more affordable. According to our analysis, over the past 18 months the cost to run a GPT-4-level model has fallen by a factor of roughly 1,000. Cheaper computation is fueling experimentation, expanding access for smaller firms and encouraging more applications.
This trend reflects a phenomenon known as the Jevons Paradox: the idea that greater efficiency leads to increased overall consumption. In the case of AI, lower costs are encouraging more companies to adopt and deploy AI solutions, which in turn is driving even greater demand for semiconductors, power and infrastructure.
Below are a few notable examples that point to a widening landscape of opportunity:
Infrastructure demand remains robust
AI’s rapid adoption is driving significant infrastructure demand. The latest AI generation models, especially those focused on reasoning and multi-step decision-making, require up to 100 times more computing power than earlier versions. This is putting sustained pressure on semiconductors, data centers, power infrastructure, and cloud networks. This also reinforces the idea that the intelligence of models is rapidly becoming very good, allowing more sophisticated use cases for AI (Figure 2).
Figure 2: The intelligence of models is increasing rapidly
Artificial analysis intelligence index
Artificial Analysis, State of AI: China, Q1 2025. 1Artificial Analysis Intelligence Index: average across a range of language model intelligence and reasoning evaluation datasets. Currently includes MMLU, GPQA Diamond, MATH-500 & HumanEval. Release date is based on first public launch of the model. 2Estimate based on company claims and comparable results where available, not yet independently benchmarked by Artificial Analysis. 3o3 Intelligence Index estimated by scaling measured Intelligence Score of o1.
Semiconductors are at the heart of the AI infrastructure story. Often referred to as the ‘picks and shovels’ of the modern world, chips power everything from cloud platforms to connected devices. As the global economy becomes increasingly tech-driven, demand for semiconductor content – across processors, power management and networking – continues to grow.
The bottom line
Rising tariffs and geopolitical friction are valid concerns. These factors have the potential to impact supply chains, increase costs for specific components or introduce other unforeseen hurdles for global technology development. Yet they are unlikely to derail the AI growth story. If anything, these challenges may reinforce the need for regional diversification and infrastructure resilience.
In a slower growth, policy-fragmented environment investors tend to reward companies with greater visibility and staying power. AI-related businesses, particularly those providing essential infrastructure or embedding AI deeply into core operations, increasingly check those boxes. From an investment perspective, rather than chasing headlines, investors should focus on identifying the enablers, early adopters and solution providers building durable advantages in an increasingly AI-powered economy.
